agentleFS
Sign inStart free
AlenroyfeildAlenroyfeild/ai-dev-workspace/template/AGENTS.md

ai-dev-workspace / template

AGENTS.md0 starsChanged 26 days agoMITView on GitHub

This folder is the shared memory for AI-assisted work on the code checkouts listed in workspace.json. Read only what the current step needs; never ingest the whole vault or repo.

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository9 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Look things up instead of re-discovering them: ws search "<words>" (product, project, runbooks, past analysis) and ws lesson search …

    line 10
  2. Very specific:

    Before changing code, think before act: is it needed, what working flows could break (gate them with a flag or version check), can existing …

    line 13
  3. Distinctive:

    Never commit or push unless the user asks in that turn.

    line 17

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

Workspace rules (all AI assistants)

This folder is the shared memory for AI-assisted work on the code checkouts listed in workspace.json. Read only what the current step needs; never ingest the whole vault or repo.

Start of a session

  1. Session-start hooks create/claim a local branch task unless capture:false; switching branches changes the brief. Without hooks, find it: ws task find <ticket-or-branch> (MCP: find_task). Read only Next action, Blockers and Handoff. No record: ws task new <ID> "<title>".
  2. Claim it before changing anything: ws claim <ID>. A claim made earlier in this workspace is resumed; a claim by another machine or person means stop and ask. One writer per checkout.
  3. Offer ws notices suggestions once; apply only after a yes. Nothing to report: say nothing. Suggest missing recommended tools once; install only after the user says yes.
  4. Look things up instead of re-discovering them: ws search "<words>" (product, project, runbooks, past analysis) and ws lesson search "<words>".

While working

  • Before changing code, think before act: is it needed, what working flows could break (gate them with a flag or version check), can existing code be reused, a numbered plan with a check per step, and proof instead of claims. Claude: /thinkbeforeact.
  • The user states a decision or correction that must hold in later sessions: call remember (CLI: ws remember "<decision>").
  • Big logs or JSON: ws digest <file> and read the summary, not the file.
  • Delegate cheap work to cheaper workers when it saves tokens; the lead assistant decides and reviews. Log each delegated step: ws run log <ID> <step> --provider <name> ....
  • Never commit or push unless the user asks in that turn. Never discard the user's edits.

Before stopping

  • ws checkpoint <ID> --status <status> --next "<exact next action>". Another assistant must be able to continue from the record alone.
  • A mistake or a useful lesson: ws lesson add "what happened → rule". When the user is annoyed by the workspace, a ws command fails, or they suggest an improvement, offer to record it: ws feedback add "..." --kind idea|bug|friction|praise (only with their yes; never include company or customer details). Sharing it with the maintainers happens later via the feedback notice.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free